DOI 10.1007/s00339-012-7402-8 I N V I T E D PA P E R
Flexible and stretchable electrodes for dielectric elastomer
actuators
Samuel Rosset· Herbert R. Shea
Received: 23 September 2012 / Accepted: 24 October 2012 / Published online: 7 November 2012 © Springer-Verlag Berlin Heidelberg 2012
Abstract Dielectric elastomer actuators (DEAs) are flexi-ble lightweight actuators that can generate strains of over 100 %. They are used in applications ranging from hap-tic feedback (mm-sized devices), to cm-scale soft robots, to meter-long blimps. DEAs consist of an electrode-elastomer-electrode stack, placed on a frame. Applying a voltage be-tween the electrodes electrostatically compresses the elas-tomer, which deforms in-plane or out-of plane depending on design. Since the electrodes are bonded to the elas-tomer, they must reliably sustain repeated very large defor-mations while remaining conductive, and without signif-icantly adding to the stiffness of the soft elastomer. The electrodes are required for electrostatic actuation, but also enable resistive and capacitive sensing of the strain, leading to self-sensing actuators. This review compares the different technologies used to make compliant electrodes for DEAs in terms of: impact on DEA device performance (speed, ef-ficiency, maximum strain), manufacturability, miniaturiza-tion, the integration of self-sensing and self-switching, and compatibility with low-voltage operation. While graphite and carbon black have been the most widely used technique in research environments, alternative methods are emerg-ing which combine compliance, conduction at over 100 % strain with better conductivity and/or ease of patternability, including microfabrication-based approaches for compliant metal thin-films, metal-polymer nano-composites, nanopar-ticle implantation, and reel-to-reel production of µm-scale patterned thin films on elastomers. Such electrodes are key to miniaturization, low-voltage operation, and widespread commercialization of DEAs.
S. Rosset (
)· H.R. SheaEcole Polytechnique Fédérale de Lausanne (EPFL), Jaquet-Droz 1, 2002 Neuchâtel, Switzerland e-mail:samuel.rosset@a3.epfl.ch
Keywords Dielectric elastomer actuators· Compliant electrodes· Carbon · Metal thin-films
1 Introduction
Dielectric Elastomer Actuators (DEAs), also known as ar-tificial muscles, are a new type of soft transducer consist-ing of a thin elastomer membrane sandwiched between two compliant electrodes. When a voltage is applied between the electrodes, the opposite charges on each electrode give rise to an electrostatic force (Maxwell stress) which squeezes the dielectric layer and causes deformation of the device. The thickness strain sz caused by the Maxwell stress is defined
by [1]:
sz= −
r0V2
Y z2 , (1)
where r and 0are, respectively, the relative and vacuum permittivity, Y is the Young modulus of the elastomer, and zthe thickness of the membrane. Driving voltages are in the order of a few kilovolts, for an elastomer membrane thick-ness of around 50 µm and with a Young modulus of about 1 MPa, leading to typical compression strains in the range of 10 %–30 % or higher.
The first report of electric field-induced deformation of a solid material is attributed to Alessandro Volta who ob-served rupture of highly charged Leyden jar capacitors in 1776 [2]. One century later, charge-induced deformation on elastomers was reported by Röntgen [3], and in 2000, Pel-rine et al. presented electrostatically activated elastomeric actuators exhibiting surface strains up to 215 % in a land-mark publication [4]. This article has marked the dawn of di-electric elastomer actuators, which quickly attracted world-wide attention.
Twelve years after its birth, the field of DEAs has made a giant leap forward, and applications are emerging in many fields and at different size scale, from ocean-wave energy harvesters to haptic feedback devices integrated in smart-phones or miniaturized devices capable of mechanically stretching single biological cells [2]. The large strains and intrinsic softness of these elastomeric actuators allow for completely different motion and systems compared to tra-ditional rigid actuators, thus allowing different approaches, such as bioinspired systems [5].
Much research effort has been invested in the selec-tion and optimizaselec-tion of the elastomers used as dielectric [1,4,6–10], as well as on applications of DEAs [11]. How-ever, much less attention has been devoted to the devices’ compliant electrodes, which are most of the time completely neglected. The fundamental deformation equation (Eq. (1)), which can be found in most of the DEA-related articles, does not take the influence of the electrodes into account and con-sequently assumes them to be infinitely thin and compliant, which they most certainly are not.
In order to enable the widespread use of DEAs, a number of challenges have yet to be solved: operation at low volt-age, miniaturization, large-scale manufacturability and inte-grated sensing to add “smartness” to the devices. The answer to these key challenges resides in better dielectric elastomer membranes, and also—though often overlooked—in opti-mized compliant electrodes. This review presents the most important technologies used to make electrodes for DEAs, and explores some new developments for electrode fabrica-tion and applicafabrica-tion that will open a broad new spectrum of applications, from high-volume production for commer-cial products, to miniaturized devices with micrometer-sized patterned electrodes.
Strictly speaking, electrodes are not absolutely necessary to induce deformation of a DEA, as shown by Keplinger et al. in a re-enactment of Röntgen’s experiment with to-day’s materials [3]. Electric charges can be directly sprayed on the elastomer’s surface by an external emitter, causing electrostatic charging and deformation of the soft dielectric due to Maxwell stress. The absence of electrodes presents some undeniable advantages: it avoids the complication of manufacturing compliant electrodes, and the operation in charge-controlled mode (as opposed to voltage-controlled in the case of the standard DEA structure) avoids the electro-static pull-in instability. Additionally, the deformation is not limited by the stiffening effect of the electrodes, or by the strain at which they lose their conductivity or break. How-ever, if spraying charges on a dielectric is relatively easy, ef-ficiently removing them is more difficult. Electrodes thus al-low to bring and remove charges quickly on the elastomer’s surface provided they are conductive enough. By patterning the electrodes with precise shapes, they can also bring the charges to precise locations, thus allowing complex struc-tures with several electrically and well defined independent
active zones on a single membrane, which is impossible in an electrode-free configuration. These two requirements are very well illustrated by the DEA-based rotary motor from Anderson et al. which combine the need for four separate electrodes on a single membrane with the necessity of high frequency actuation [12]. Electrodes of DEAs are a there-fore a practical necessity, and care must be taken in their de-sign in order to profit from their ability to efficiently move the charges at precise locations, without too much impact on the stiffness of the actuator.
2 Stretchable electrodes for soft machines and conformable electronics
Although located on the actuator’s surface, the electrodes of dielectric elastomer actuators are at the core of the de-vice’s performance: they must be conductive, yet they must be soft; they must sustain large deformations while remain-ing conductive, yet they must be able to do so for millions of cycles. The life of a compliant electrode for DEAs is a tough one, and it comes as no surprise that the ideal elec-trode has yet to be developed. However, much progress has been made from the first reported devices whose electrodes were hand-painted carbon grease, and over the years, cre-ative new methods and technologies have started to emerge. Compliant electrodes are needed in a wide range of ap-plications, well beyond DEAs. Indeed, the booming field of soft and deformable electronics also relies on flexible inter-connects that can sustain stretching and/or bending, as re-cently highlighted in the march 2012 special issue of the MRS Bulletin [13]. The requirements regarding strain am-plitude, lifetime and conductivity is quite different between DEAs and stretchable electronics: the electrodes of a DEA must typically sustain biaxial strains of 10 %–100 %, while the deformation of soft electronic circuits is often limited to about 20 % uniaxial or bending strain. An electrode which is highly stretchable will also be bendable, while a bend-able electrode is not necessarily highly deformbend-able. DEA must typically survive a large number of stretching cycles (>50 millions cycles in the case of an ocean-wave energy harvester (cf. Sect. 6.4)), while in the case of stretchable electronics it is generally lower and varies from a single deformation in the case of an implantable sensor to sev-eral thousand cycles in the case of a wearable flexible dis-plays. Regarding electrical conductivity, flexible electronics require electrodes with a good conductivity to avoid volt-age drop along the conductive tracks, while DEAs, being electrostatic devices can tolerate electrodes with a much higher surface resistance, although a good conductivity is needed for high speed operation. Despite these differences, electrodes for DEAs and stretchable electronics have many common characteristics, such as materials, deposition meth-ods and patterning solutions. Consequently, the applications
Fig. 1 Application of deformable electrodes: (a) Electronic artificial skin with integrated organic transistors and pressure sensors. From [15], ©2012 Materials Research Society. Reproduced with permis-sion. (b) Stretchable and deformable interconnects between two PCBs. From [16], ©2012 Materials Research Society. Reproduced with per-mission. (c) Deformable electronic circuit applied on the backside of a temporary tattoo. The circuit and tattoo can conform to the skin and sustain deformation. From [17], ©2011 AAAS. Reproduced with per-mission from AAAS. d) Dielectric elastomer-based wave energy con-verter from the company SBM offshore. From [18], ©2012 SPIE. Re-produced with permission. (e) DEA-based ring oscillator with smart self-sensing and self-switching integrated on the compliant electrodes. From [19]
of the soft electrodes technologies presented in this review largely exceed the sole domain of DEAs, but encompasses bioinspired stretchable electronics [14], soft electronic skins [15] or PCB-inspired stretchable circuits [16] to give a few examples (Fig.1).
The electrodes of a DEA are subject to the same amount of deformation as the polymer and must consequently sus-tain strains typically between 10 %–100 % without being damaged and while remaining conductive. Their mechani-cal impact on the stiffness of the dielectric must be low to avoid reducing the strain output. In one word, they must be compliant. As stated by Pelrine et al.: “The ideal elec-trode would be highly conductive, perfectly compliant and patternable, and could be made thin relative to the poly-mer thickness” [1]. The general compromise found in the first EAP-related publications is to use electrodes based on carbon powders applied with a brush or a spray, which— except for the compliance—are far from the perfect elec-trodes described by Pelrine. Graphite and carbon black pow-ders, conductive carbon grease, or carbon powders in an elastomeric matrix were the first types of electrode used, because of their low stiffness and ability to remain con-ductive at large strains [1,4,20,21]. Although well-adapted to lab experimentations, carbon-based compliant electrodes applied with a paintbrush have limitations in terms of pat-ternability, scalability for large-volume production, compat-ibility with clean-room processes, reliability, lifetime and
ease of application. To overcome these limitations, new types of electrode have emerged over the last few years. Metallic thin films, although not intrinsically stretchable be-yond a few percent, can be patterned in-plane or made to ripple out-of-plane, and can then remain conductive and un-damaged at strains above 30 %. Other methods including nano-composites are also emerging, based on metal ion im-plantation, platinum salts, exfoliated graphite or carbon nan-otubes.
In the following sections, the different compliant elec-trodes used for DEAs will be presented, starting with the most widely used carbon-based electrodes (Sect. 3), fol-lowed by the metal thin-film electrodes (Sect. 4), and fi-nally by some emerging more exotic methods (Sect.5). The main characteristics of each electrode type are investigated, as well as the methods available to pattern these electrodes on thin elastomeric substrates. In the discussion (Sect.6), several practical applications (miniaturization, reduction of driving voltage, integrating sensing and energy harvesting) are presented; the impact of the electrodes for each of these examples is examined in detail.
3 Carbon-based electrodes
3.1 Carbon powder, grease, and rubber electrodes
The most commonly used electrodes for DEAs are based on carbon particles and can be categorized in three main vari-ants (Fig.2): loose particles of carbon simply deposited on the elastomeric membrane; carbon grease, which consists of carbon particles dispersed into a viscous media such as oil; and conductive rubber, which is formed by dispersing car-bon particles into an elastomer, which is crosslinked after it has been applied on the membrane.
Carbon powder One of the main advantages of powder-based electrodes is that they do not contribute to the stiff-ness of the membrane on which they are applied, due to the absence of a strong binding force between the agglomerates. Loose carbon powders (mainly carbon black and graphite), directly applied on the dielectric membrane, were thus the material of choice in the early days of DEAs, as they allow demonstrating the large strain capabilities of these soft ac-tuators. Although metals are intrinsically more conductive than carbon, their powders have a tendency to form an in-sulating oxide layer at the surface [23]. Additionally metal particles are too large to create sufficient contacts, and a bet-ter conductivity can be obtained with carbon particles [24]. Metallic powders are therefore rarely used for DEAs, al-though a few articles report on the use of silver grease elec-trodes [25,26]. The relatively high electrical resistance of carbon-based electrodes does not play a primary role for
Fig. 2 The three main types of carbon-based electrode. (a) Loose carbon powders consists of carbon black (or graphite) particles directly applied on the elastomer membrane. (b) Carbon grease consists of carbon particles dispersed in a viscous oil. (c) Conductive rubber consists of carbon particles dispersed into a crosslinked elastomer. Photos adapted from [22], ©2007 SPIE. Reproduced with permission
actuators, as long as the electrical time constant of the de-vice remains smaller than the mechanical bandwidth. For in-stance a 1 cm2, 40 µm thick DEA with a silicone membrane has a capacitance of about 65 pF. With electrodes whose re-sistance would be in the M range, it could still be driven up to 1 kHz.
In addition to being quite difficult to handle due to their high sensitivity to static charges, loose powders present sev-eral disadvantages: maintaining full coverage at large strains is difficult, [9,27] and the lifetime of the electrode is lim-ited due to the possibility of the conductive particles to de-tach from the electrode. Consequently, electrodes with loose powders are mainly found in combination with adhesive acrylic elastomers (such as VHB from 3M) as dielectric layer: because this material is intrinsically sticky, the con-ductive particles become bound to its surface. A notable ex-ception where loose graphite powder electrodes are used in combination with a silicone elastomer is in the multilayer process of Prof. Schlaak’s group [28–31]. But as each elec-trode is effectively covered by the next dielectric layer, one cannot strictly speak of loose powders for this particular case.
Carbon grease One way to solve the issues mentioned above is to bind the powders into a viscous matrix, such as grease. Although easier to handle than loose powders and capable of sustaining larger strains while remaining con-ductive, carbon-grease electrodes also have disadvantages: grease can have long term stability issues caused by drying or diffusion into the dielectric membrane, which can cause short circuits or swelling of the elastomer membrane. As any viscous material, grease also creeps under gravity, which negatively impacts the lifetime of these electrodes, particu-larly for devices stored vertically. The absence of a reliabil-ity study on these kinds of electrode (and DEAs in general) makes it difficult to assess the importance of these factors. Furthermore, grease, similarly to loose powder, can be sub-ject to mechanical abrasion, which also negatively impacts the lifetime.
Conductive rubber Finally, the conductive carbon black particles can be incorporated into an elastomer matrix, such as a silicone, which is cured after the electrode is applied on the membrane, in order to obtain a polymer-carbon con-ductive composite. Because in that case the electrode is bound to the membrane, it is much less prone to ablation or migration of the electrode material with time, which pos-itively impacts the electrode’s life expectancy. However, be-cause of the elastomeric matrix, the electrode’s contribu-tion to the stiffness of the elastomer cannot be neglected anymore, compared to carbon powder or grease. As Pel-rine et al. pointed out, these types of electrode work best for thicker dielectric films, i.e. when the thickness of the electrode is negligible in comparison with the thickness of the dielectric elastomer [1]. The stiffness of an elastomer-carbon black compound is very dependent on the quantity of filler particles, which must be sufficiently high to be above the percolation threshold. The percolation threshold is very dependent on the surface area of the carbon black, as well as the matrix in which it is dispersed and can vary between 1 %–24 % [23]. In silicone, we have observed a percolation threshold of about 6 % for Ketjenblack EC-300J from Ak-zoNobel, and of about 3 % for Ketjenblack EC-600JD, due to its extremely high surface area (1400 m2/g according to the datasheet, compared to 800 m2/g for the former)1. With graphite (4206 from Merck) in DowCorning Sylgard 184, Kofod observed a percolation threshold of 23 % [32].
Assuming the same silicone is used to produce the di-electric membrane and the conductive electrodes, the me-chanical stiffness of the filled polymer will be larger than the membrane’s, thus showing the importance to apply the material in a very thin layer.
The main carbon-based compounds reported in literature for the fabrication of compliant electrodes for DEAs are summarized in Tables1,2and3. In addition to these three
main types of carbon electrode for DEAs, some other ideas have been presented, for example using carbon nanotubes (cf. Sect.5.3) or exfoliated graphite [39].
Table 1 Main carbon powder-based compounds used to make elec-trodes for DEAs
Type Name Ref.
Carbon black Vulcan XC72 [32,33]
Carbon black Ketjenblack EC-300J [32,34,35] Carbon black Ketjenblack EC-600JD [36]
Graphite N-77 (Spray) [20]
Graphite Merck 4206 [32]
Table 2 Main carbon grease-based compounds used to make elec-trodes for DEAs.∗indicates a compound prepared in the research lab-oratory
Name Ref.
CW 7200, Chemtronics [9,20]
Nyogel 755G, Nye Lubricants [37]
Nyogel 756G, Nye Lubricants [19]
PDMS oil+ Ketjenblack EC-300J∗ [32]
Table 3 Main conductive rubber-based compounds used to make elec-trodes for DEAs.∗indicates a compound prepared in the research lab-oratory
Composition Ref.
RTV 60-CON, Stockwell [9,25]
Rhodorsil CAF4, Bluestar+ Vulcan XC72∗ [33]
RTV23, Altropol+ KB EC-300J∗ [35]
Elastosil E43, Wacker+ KB EC-300J∗ [32] Sylgard 184, DowCorning+ Vulcan XC72∗ [38]
3.2 Application methods for carbon electrodes
While simple prototypes and demonstrators can be manu-factured without much regard of the shape, precision and thickness homogeneity of the electrodes, which can simply be smeared on the elastomeric membrane [19,21,37,40], most actuators for commercial applications require elec-trodes of a precise shape, patterned on a cm, mm, or even µm scale. Several techniques can be used to apply and pat-tern carbon electrodes on a dielectric elastomer (Fig.3), as described in the next paragraphs.
Shadow masking A shadow mask, or stencil can be placed on the elastomer membrane to selectively expose the surface that needs to be coated with the conductive material [1]. If the thickness homogeneity and reproducibility is not a main concern, the conductive solution can be simply applied on the mask with a brush, which has been demonstrated by Pelrine et al. for loose powders and carbon grease [9]. For better uniformity, spray coating can be used in combination with a shadow mask to apply a thin uniform layer on the elastomeric membrane. For their automatized stack process, Schlaak et al. have tested spray coating through a shadow mask of both dry graphite powder and graphite in suspen-sion in isopropanol [28–31,41]. With their process, which alternates between the application of a silicone layer by spin coating, the heat-activated cross-linking of the silicone, and the subsequent deposition of the electrodes (Fig. 4), they have developed an efficient production method, which al-lows building actuators of up to 100 layers with a production time of 5 minutes per layer, thus demonstrating that carbon electrodes can also be used for larger scale and commercial applications. Araromi et al. reported on a similar stacked actuator fabrication process with a sequence of silicone and electrode deposition by spray coating with an airbrush [42]. Because the dielectric layer is also sprayed, there is no need for a spin-coater, thus reducing the cost of the installation.
Fig. 3 Different ways to pattern carbon electrodes. (a) Using a shadow mask to selectively protect part of the elastomeric membrane. The carbon-based electrode material can then be dispensed (for example by spraying) on the surface. The shadow mask is subsequently removed to
expose the patterned electrode. (b) Using a patterned elastomeric stamp to pick-up the electrode material and apply it on the elastomeric mem-brane. (c) Using standard printing techniques, such as drop-on-demand inkjet printing
Fig. 4 Main process steps of the Schlaak et al. process for manufac-turing stacked DEAs with patterned carbon sprayed electrodes. From [31], ©2011 IEEE. Reproduced with permission
Their electrode consists of graphite powder put in suspen-sion in a solvent (1:4 ratio). Spray coating through a shadow mask can also be used with conductive rubber if mixed with solvent to decrease viscosity [32,35].
One drawback of using a stencil to define the shape of the electrodes is the contact between the mask and the elas-tomer membrane, particularly in the case of very thin sus-pended membranes. If applying the mask and spraying the electrode can be done quickly, removing the mask at the end of the process must be done slowly and carefully in order not to damage the membrane stuck on the mask. Leaving a small gap between the mask and the membrane is not advis-able because shadowing will occur at the mask border and the airflow from the spray coating system can deform the membrane, leading to a loss of resolution.
Stamping Patterned structures can also be obtained by stamping the conductive electrode on the dielectric mem-brane. A soft stamp with the desired pattern is fabricated, for example by replication on an etched silicon negative master. Small structures and a good resolution can be obtained with this method. For instance, Aschwanden and Stemmer have used this technique to pattern 100 µm-wide lines with loose carbon black powder on an acrylic elastomer (VHB 9460) membrane [43]. The lines of electrodes on both sides of the membrane formed a matrix of 100 µm× 100 µm actuators which exhibited up to 35 % linear strain for a 500 V actu-ation voltage (Fig.5). Stamping loose carbon black powder is made possible for this particular application by the use of the VHB adhesive as dielectric layer: the stickiness of the surface ensures the transfer of the carbon particles from the stamp to the membrane, and fixes them on the elastomer sur-face. Commercially available stamping techniques can also be used with carbon grease or conductive rubber when ade-quately diluted to form a conductive ink.
Printing techniques There is still room for development of effective, rapid and high-resolution patterning techniques for carbon-based electrodes for DEAs. The three major
Fig. 5 100 µm-wide loose carbon black electrodes applied with a structured stamp on an acrylic elastomer membrane. From [43], ©2007 SPIE. Reproduced with permission
types of carbon electrode presented here can be prepared in the form of a conductive ink, and most of the techniques developed for the printing industry, including inkjet print-ing, screen printing or roll-to-roll processes, can be applied to pattern conductive electrodes on elastomers. The poly-mer solar cell community, for example, is using these print-ing methods for the manufacturprint-ing of flexible devices [44]. Their production lines require high throughput, the coat-ing of large surfaces, and the patterncoat-ing of precise features (down to the µm scale). This is similar to the needs of com-pliant electrodes for DEAs in case of a large-scale commer-cial application.
Among the standard printing methods, inkjet printing is particularly interesting, because it is a purely non-contact method, and it is therefore well-adapted for thin suspended membranes. This is a versatile method, as the shape of the electrode can be adapted at will just by modifying the bitmap used to print the electrode. This represents a very interesting advantage over shadow masks for the research community, because different designs can quickly be printed and tested, which allows for quick optimization of the electrode shape. Large-scale printers with multi-nozzle (up to 1000) print-heads are capable of large throughput, which renders this technique also attractive for commercial high-volume pro-duction of DEAs. A major difficulty resides in the devel-opment of a jettable ink for EAP applications, which is not a straightforward task. The parameter space is limited by a large number of factors: first, there are the printer/nozzle requirements on liquid viscosity, surface tension and vapor pressure, which must be followed in order to obtain regular and stable generation of drops on demand. These require-ments are nozzle-related and can be quite tight. For example, the R&D printer Dimatix DMP-2800 from Fujifilm requires inks with viscosities between 10 and 12 mPa s, and surface tension between 22 and 33 mN/m. To avoid clogging the small nozzle orifices (typically 10–100 µm depending on manufacturer), the carbon particles must be carefully dis-persed and no flocculation or sedimentation is allowed. The evaporation rate of the ink must not be too high, to avoid
Fig. 6 Schematic representation of the reduced parameter space avail-able to design an inkjettavail-able ink for DEAs application. The fluid must fulfill the requirement of the printer’s manufacturer in order to be re-liably dispensed. Once the ink droplets reaches the polymer’s surface, it must cover it and dry uniformly, and finally the cured ink must be conductive and compliant in order to be used for DEAs
clogging the nozzle when it is not jetting (i.e. when the noz-zle is traveling above a zone which must remain free from ink). Secondly, the interaction between the droplet and the membrane, once the former reaches the latter, adds further requirements on the fluid formulation. The contact angle be-tween the ink and the elastomer must be low enough to ob-tain good wetting. If the contact angle is too high, there is a risk that adjacent droplets merge together and form a large ink pool, causing a dramatic loss of resolution. This is espe-cially problematic with silicones, which have a very low sur-face energy and present highly hydrophobic sursur-faces. Differ-ent surface treatmDiffer-ents (plasma activation, UV exposure) can help to temporarily increase the surface energy, and Robin-son et al. have recently demonstrated how the wetting of a silicone substrate by a solvent-based ink can be improved by structuring the silicone substrate with micropillars [45]. The ink’s evaporation rate must be controlled so as to avoid the “coffee stain” effect, which is caused by the motion of the ink’s solid content to the periphery of the droplet, leading to a non-uniform coating once the layer is dry. This effect can be avoided by carefully choosing the solvents used for the ink formulation, using a mix of solvents with different va-por pressures [46]. Finally, the ink formulation must be se-lected so that the dry and cured film fulfils the requirement for compliant electrodes: low stiffness, compliance, conduc-tivity etc. Figure6illustrates the requirement that an ink for-mulation must fulfil.
3.3 Electrical properties of carbon electrodes
Because of the absence of suitable commercial products, most carbon-based compounds are prepared in research lab-oratories, with different amount of fillers, different degrees
of dispersion of the agglomerates, and applied in a broad range of thicknesses. Consequently, a large spectrum of sur-face resistance has been reported in the literature. Kornbluh et al. commented on this large resistivity spectrum by stat-ing: “Typical surface resistivities of our electrodes are on the order of tens to hundreds to thousands of ohm” [8]. Toth and Goldenberg measured a surface resistance of about 20 k / for carbon grease. They observed an exponential increase of the electrodes resistance when submitted to an uniaxial strain, which they modeled with the following equa-tion [25]:
R= R0β(α−1), (2)
where R is the resistance of the stretched electrode, R0the initial resistance of the undeformed electrode, α the uniax-ial stretch, and β a constant characterizing the sensitivity of the resistance to strain. They tested different types of elec-trode and reported the following values for β: 4 for car-bon conducting grease, 6.5 for conducting RTV silicone, 6.8 for silver conducting grease, and 22 for graphite pow-der. Carbon-grease electrodes was thus expected to remain conductive at larger strain compared to the other electrodes tested in the study. Carbon grease is thus the electrode ma-terial used by Keplinger et al. to demonstrate gigantic strain (1692 % surface strain) on a DEA balloon, using instability snap-through [47]. However, with another home-made car-bon grease, Kofod observed that grease migration starts to occur at large strains (above 100 % biaxial), and full cov-erage of the electrode surface is not ensured anymore [32]. This highlights how the electrical properties (initial resistiv-ity and behavior when strained) of home-made carbon-based compounds depend on the method of preparation.
In addition to their primary role of transporting and stor-ing the charges for the generation of the electrostatic force (and hence actuation), the electrodes, and their change in resistance while being deformed can be used to monitor the strain and operate the actuator in closed-loop mode. O’Brien et al. characterized the three main types of carbon-based electrode (loose powder, grease and conductive rubber) for resistive self-sensing (Fig.2) [22]. For the three electrode types, they obtained a surface resistance of about 9 k /. The electrodes were then submitted to uniaxial strain, while measuring the change in resistance for self-sensing appli-cations. Loose carbon powder exhibited the best resistance as function of strain behavior: no noise, and no overshoot, while carbon grease produced a noisy signal and bound car-bon powder—although mechanically the best—presented a time-dependent signal with strong overshoot at each change in strain rate (Fig.7). As explained by Wang et al., the resis-tance time-dependency of a carbon-filled elastomer exhibit the same forms as the corresponding mechanical properties of a viscoelastic material, thus explaining the jumps in resis-tance at the times at which the strain rate is modified [48].
Fig. 7 Change in resistance (normalized to the initial track length) of carbon-based compliant electrodes submitted to uniaxial stretching. (a) Loose carbon powder, (b) conductive rubber and (c) carbon grease. (d) Normalized resistance as a function of strain for the three types of electrode. Adapted from [22]
Fig. 8 Left: example of interface between a compliant electrode and an electronic circuit with the help of a double-sided PCB playing the double role of a support frame and electrical interface through the use of metallized vias. Right: application of the concept to a DEA with a home-made single-sided PCB. The absence of copper on the backside (in contact with the electrode) is compensated with a conductive silver varnish in the via hole
The plot of the resistance as a function of the applied strain shows the lack of hysteresis of the loose carbon electrodes compared to the two other alternative, which is very impor-tant for self-sensing applications (Fig.7(d)). One also notes the higher sensitivity of the resistance of loose carbon pow-der electrode compared to grease, which is in good agree-ment with the empirical relation and observations of Toth and Goldenberg reported in the previous paragraph.
Compliant electrodes are necessary to obtain a large dis-placement of the active area (the zone which is covered by two electrodes). In order to bring the electrical charges that will create the electrostatic force and the displacement of the active zone, it is necessary to connect one end of the elec-trode to a high-voltage power supply. The interface between the electronic driving circuit and the soft electrodes must be made in a way to ensure a reliable electrical contact with-out damaging the electrodes and while assuring a long life-time of the interconnect. Because of their crosslinked state and good adhesion to the dielectric elastomer layer on which they are applied, conductive rubber electrodes are easier to reliably connect to an external rigid electrical connection point compared to loose powder or carbon grease. There are several methods to connect an electrode to the driving cir-cuit. One possibility in the case of suspended membranes is to combine the supporting frame holding the membrane with the electrical contact interface, by using an insulating struc-ture with conductive tracks, such as a printed circuit board (PCB). It is then possible to integrate the actuator with the electrical driving circuit on the same board (Fig.8).
4 Metallic thin-film electrodes
The microelectronics and MEMS industry has developed a broad range of microfabrication technologies to create
pat-terned conductive thin films. Metals can be deposited in thin layers on a broad range of substrates by electron beam evap-oration, cathodic sputtering, or electroplating and patterned down to the nm scale using photolithographic processes. Re-producible results at a large throughput can be obtained with these methods (for instance over one billion interconnects on modern microprocessors), which consequently have the po-tential of moving the EAP technology from the lab to the in-dustry. The ability to pattern the electrodes on a small scale allows for the fabrication of many electrically independent actuators on the same membrane, which paves the way for a broad range of new applications based on arrays of microac-tuators.
However, there are two major obstacles to the direct use of metal thin film as electrodes for DEAs. First, the Young modulus of metal is several order of magnitude higher than that of dielectric elastomers (50–100 GPa compared to 0.2–1 MPa), which means that even 50 nm-thick electrodes on a 50 µm-thick elastomer will have a significant stiffen-ing impact on the elastomer, leadstiffen-ing to a negligible actua-tion strain. For example, we have shown that sputtering a 8 nm gold layer on a 30.6 µm PDMS membrane with an ini-tial Young’s modulus of 0.77 MPa caused a relative increase of the Young modulus of the membrane of 440 %, up to 4.2 MPa [49]. Secondly, the limit of elasticity for metals is very low, typically 2–3 % and if a metal electrode is strained above this limit, it will crack and form islands separated by non-conductive polymer.
Despite this limitation some research groups proposed and tested clever solutions to use metallic thin films for DEA applications, even though they present a very high Young’s modulus and a low limit of elasticity. These methods, sum-marized in Fig. 9, include the use of patterned electrodes (Sect. 4.1), of out-of-plane buckled electrodes (Sect. 4.2), and of corrugated membranes (Sect.4.3).
4.1 Patterned metal electrodes
Patterning metal traces can be used to define several inde-pendent devices on the same membrane, but on a smaller scale, it can also be advantageously used to replace a sheet (i.e. continuous) electrode by a patterned one (Fig. 10). When carefully designed, the patterned electrode still allows for the membrane to move and expand in the desired direc-tion without damaging the metallic thin film. Compared to a plain electrode with the same external dimension, the pat-terned electrode has a smaller active area, and therefore the capacitance formed by two aligned electrodes between an elastomeric membrane is lower, leading to a smaller elec-trostatic force and strain. However, if the patterned structure is carefully designed, the capacitance can be increased due to the fringing field capacitance, and approach that of plain electrodes.
Fig. 9 Different methods to create stretchable electrodes based on metal thin films. (a) Using a patterned electrode defined by pho-tolithography instead of a plain electrode. (b) Depositing the metal-lic thin film on a stretched elastomeric membrane, in order to create out-of-plane buckling when the substrate is relaxed. (c) Fabricating a corrugated elastomeric membrane on which to deposit the thin-film electrode
Fig. 10 Patterned compliant metal thin-film electrodes. (a) 170 nm Cr/Au/Cr electrodes evaporated on a 30–40 µm-thick DowCorning Sylgard 186 membrane and patterned to optimize out-of-plane dis-placement of the diaphragm. Adapted from [50], ©2007 IEEE. Repro-duced with permission. (b) Horseshoe-shaped metal tracks optimized for uniaxial extension. From [51], ©2012 IOP publishing Ltd. Repro-duced with permission
Pelrine et al. have deposited a thin layer of gold (<100 nm) on an elastomeric membrane by cathodic sput-tering. The layer was subsequently patterned by pho-tolithography to form zig-zags with a conductive width down to 5 µm [9,27]. They report that the electrodes can be stretched in the direction of the zig-zags up to 80 % while remaining conductive. It should be pointed out that the mechanical properties of a membrane with such zig-zag electrodes patterned on its surface become anisotropic, and that the membrane is softer in the direction of the zig-zags than across them. When used for DEAs these electrodes promote uniaxial strain, which can, depending on the practi-cal application in which the actuator is intended to be used, be seen as an advantage or a disadvantage. To reduce the stiffening impact on the membrane, the zig-zag electrodes of Pelrine et al. are separated by a space many times their width. Consequently, to ensure a better electric field cover-age, the authors point out that a second electrode material is necessary to carry the charges in-between the zig-zag traces, and they have shown that the humidity contained in ambi-ent air makes it sufficiambi-ently conductive to conduct charges between two gold traces, if speed is not an issue.
Although not directly targeted at DEAs, Gonzalez et al. have studied the maximal elongation of meandering metal-lic tracks deposited on a simetal-licone elastomer and submitted to uniaxial strain (Fig.10(right)) [52]. The authors have con-ducted FEM simulations to optimize the shape of the me-anders in order to maximize the elongation of the conduc-tive track at break. They found that a “horseshoe” shaped track with a conductive width as small as possible was the best solution, and they reported elongation at break up to 100 %, which is slightly above the non-optimized zig-zags tracks of Perlrine et al. presented in the previous paragraph. The technique was further developed by Verplancke et al. by sandwiching the meandering track between two spin-coated polyimide layers as a means to improve mechanical perfor-mance [51]. They obtained reproducible uniaxial stretch-ing up to 100 % without noticeable change in resistance (R0≈ 180 ) or hysteresis. A life time of 500000 stretching cycles has been demonstrated for a strain of 10 %.
Pimpin et al. have designed a diaphragm microactuator capable of out-of-plane buckling made with a 180 nm gold electrode deposited by e-beam evaporation. A 10 nm chromium adhesion layer was deposited first, followed by the gold layer and a final chromium layer to avoid stress gra-dients. The metallic layer was finally patterned in concentric rings to form a compliant electrode [50,53]. This example is particularly interesting, because it presents a small-size de-vice (2 mm diameter) made in a clean-room environment with microfabrication production methods. Carbon-based electrodes are consequently out of question for this pro-cess, and using electrodes patterned in concentric rings is a clever workaround. To obtain the best strain performance,
the electrodes must be carefully designed. For example in-creasing the spacing between the conductive tracks while keeping the width of these lines constant reduces the stiffen-ing effect of the metal, but reduces the effective area of the electrode. Pimpin et al. have conducted FEM simulations and experimental verifications to optimize the fill factor of their patterned electrode. Taking the width of the conduc-tive tracks w, the gap between the concentric rings g, and the thickness of the dielectric membrane h, they found that the displacement was maximized when the electrodes have a g/w ratio of 0.33, and a w/ h ratio as small as possible. Consequently, g should be made as narrow as the pattern-ing process allows, and w three times larger than g. With g= 7 µm and w = 23 µm, the authors obtained a static de-flection of 112 µm at an applied electric field of 75 V/µm, which represents 5.6 % of the membrane’s diameter. The authors also tested unpatterned electrodes of the same total size and obtained a static displacement six times smaller, thus showing the necessity of patterning a metallic thin-film electrode when used for DEAs. However, even with the patterned electrode, the 5.6 % height over diameter ra-tio is not a large strain: assuming the membrane’s buckled shape is a spherical cap, this vertical displacement corre-sponds to a surface strain of the membrane of 0.31 %. On the same elastomer (Dow Corning Sylgard 186) but with ex-tremely compliant electrodes made with carbon grease, we have measured surface strains of 14.8 % at the same applied field.2
4.2 Out-of-plane buckled electrodes
The high coefficient of thermal expansion of silicones (9× 10−4K−1(volume) for Dow Corning Sylgard 186 according to the datasheet) is often problematic when they are used in conjunction with other materials with much lower thermal expansion. However, this property can be advantageously used to make compliant electrodes out-of-plain metallic thin films. In a letter to Nature, Bowden et al. showed how they were capable of creating complex out-of-plane structures by depositing a thin-metallic film (50 nm) by E-beam evapora-tion on a piece of heated silicone (Fig.11) [54]. The con-traction of the silicone when it was cooled down to room temperature created a compressive stress in the metallic thin film, causing it to buckle out-of-plane, forming an undulat-ing structure with a wavelength of 20–50 µm. These kind of structures can be induced either by heating the silicone substrate, or just by using the heat generated by the de-position process. Lacour and Wagner evaporated 100 nm-thick gold tracks that were 28 mm long and 0.25 mm wide on a 1 mm-thick silicone substrate and observed the for-mation of out-of-plane structures perpendicular to the track
Fig. 11 Out-of-plane buckling of a thin-metallic film deposited on an elastomer membrane due to different coefficient of thermal expansion. From [54], ©1998 Macmillan Publishers Ltd. Reproduced with per-mission
length [55]. They measured the resistance increase of the track when submitted to uniaxial strain, and observed that it remained conductive for strains up to 23 %, which greatly exceeds the yield strain of gold. This is due to the fact that the out-of-plane waves flatten out when the composite is stretched. Similar out-of-plane wavy electrodes induced by thermal stress have also been reported by Maghribi et al. for a stretchable micro-electrode array [56].
Controlling the amplitude of the out-of-plane structures is difficult if one relies uniquely on the thermal stress in-duced during the deposition. However, generating compres-sive stress in the mechanical layer is also possible by pre-stretching the elastomeric substrate before the deposition of the metallic thin film (Fig.9(b)) [55,57]. Lacour et al. have uniaxially prestretched a silicone substrate by 10 %–20 % before depositing a 25 nm-thick gold layer [55]. Following deposition and patterning of the metallic tracks, the substrate was allowed to relax, leading to out-of-plane buckling of the gold electrode. The resistance of 0.5 mm × 4.6 mm stripes was measured as a function of strain. Electrodes that were evaporated on a substrate that was stretched 15 % during the deposition process could be uniaxially strained up to 28 % before losing conductivity because of crack formation in the metallic layer. It should be pointed out that this value is far superior to the 15 % strain state applied during the electrode deposition, strain at which the out-of-plane waves are ex-pected to completely disappear when subsequently stretched in the same direction. After the release of the 15 % deposi-tion prestretch, the metallic stripes formed a sinusoidal pro-file with wavelength of 8.4 µm and an amplitude of 1.2 µm.
The authors report on a large variation of the initial sur-face resistance of the stripes, from 0.8 / up to several M / [55]. The sample whose resistance was character-ized as function of strain had an initial surface resistance of ∼14 /. Upon uniaxial stretching, the resistance of the stripe decreased down to∼8.2 /, for a strain of 17 %, i.e. slightly higher that the deposition prestretch. At higher strain, the surface resistance increases slowly. Above 28 % strain it rises sharply up to electrical failure. Cyclic testing of up to 100 cycles have shown that the samples remain sta-ble. To apply this kind of electrode to DEAs, however, the cyclic loading tests should be pushed further. This method can also be used with a biaxial prestretching for deposition, in order to make actuators capable of expanding in the two in-plane directions, depending on the application.
Electrodes which remain conductive when stretched are one of the two important requirements for compliant elec-trodes for DEAs, the second being a negligible stiffening impact on the dielectric membrane. This effect has not been analysed for these buckled electrodes, and they have not been directly applied to build and characterize DEAs. How-ever, their close similarity with the corrugated electrodes (cf. Sect.4.3) which have been successfully used to make reliable DEAs, indicates that similar performance should be expected.
4.3 Metal thin-film electrodes on corrugated membranes
One of the main limitations of the method presented in the previous paragraph is the need for mechanically prestretch-ing the membrane. Metal thin-film deposition and pattern-ing via photolithography are methods which are compat-ible with a microfabrication process flow, but prestretch-ing is not, due to the need to remove the membrane from its substrate, mechanically deform it and fix it on a retain-ing frame. As an alternative method to mechanical or heat-induced prestretch to generate out-of-plane wavy electrodes, Benslimane et al. have introduced the use of a structured di-electric elastomer membrane [58]. The silicone elastomer is applied on a mold with a quasi-sinusoidal corrugation pro-file. After curing, the membrane is separated from the mold, and a 70–110 nm thick Ag thin film is deposited on the cor-rugated side. The wave profile of the mold has been opti-mized to obtain elongations of 33 % in the compliant di-rection before the metallic electrode breaks. The capacitor configuration needed for DEA applications is obtained by laminating two films together with their flat sides in contact (Fig.12). The lifetime of the corrugated electrodes was in-vestigated by cyclic activation of an actuator with corrugated electrodes. The test setup consisted of two membrane actua-tors mounted in a push-pull configuration. One membrane was activated, and the other acted as a counter-balancing spring. The actuation signal was 2.5 kV at 30 Hz, and more
than 3 million cycles were demonstrated without change in the actuation strain or damage to the electrodes [58].
This technology has been developed by the Danfoss group in Denmark in collaboration with RISØ DTU. In July 2008 a separate entity, Danfoss Polypower A/S,3has been founded to further develop this process, and to manufac-ture DEAs with corrugated electrodes on a large scale. Dan-foss Polypower were the first to manufacture in high vol-ume an electrode-coated DEA material. They have devel-oped custom machinery that enable roll-to-roll production of hundreds of meters of DEA membranes, from the corru-gated membrane fabrication to the lamination of two films to form a capacitor [59]. The electrodes are deposited by a roll-to-roll vacuum sputtering process, which follows the delamination of the membrane from the mould, whose pro-file height and period are respectively 4 µm and 10 µm. Plasma treatment is used to promote the adhesion of the metal on the elastomer prior to the metallization. The infras-tructure allows the coating of different metals, and a roll-to-roll masking mechanism can be used to pattern the elec-trode, if needed. Recent research effort for the optimization of the corrugation profile has lead to metallized membranes capable of sustaining a strain of 80 % without inducing dam-ages to the metallic layer, by using a mold profile with a 1:1 height over period ratio [60]. The authors have computed an
Fig. 12 Corrugated membrane with thin-film metallization applied on the corrugated surface. The sandwiched configuration necessary to make dielectric elastomer actuators is obtained by laminating two met-allized membranes back to back
3http://www.polypower.com.
experimental “compliance factor” describing the stiffness of the coated membrane in the compliant direction, and esti-mated that the new profile is about 4.5 times more compliant than the previous one, which should lead to a strain gain for the same applied electric field when use in a DEA config-uration. According to their model, the corrugated metallic electrode with this new profile is 8100 times more compli-ant than the same metallic film deposited on a non-structured (i.e. flat) membrane.
Push actuators named InLastor were fabricated by rolling the DEA membrane in a cylinder of many layers. With 2500 V actuation (∼31 V/µm), the actuators exhibit a 3 % linear strain, a blocking force of 6 N/cm2 and a response time below 10 ms [59].
4.4 Patterning methods for metallic thin-film electrodes
One of the main advantage of using metallic thin film as electrode material is the ability to precisely define the shape of the electrode via different methods of patterning (Fig.13). The three techniques compatible with elastomer membranes are as follows.
Shadow masking A shadow mask can be used to selec-tively expose the surface of the elastomer during the coating process, similar to what can be done with carbon-based elec-trodes. Laser cutting of thin (<100 µm) steel sheets can be used to obtain structures down to∼100 µm. Smaller struc-tures, down to a few micrometers, can be achieved through the use of a microfabricated stencil [61]. Using focused ion-beam milling for the fabrication of the stencil, deposition of metallic structures smaller than 100 nm has been demon-strated [62]. But the fabrication process of nano-stencil is quite complicated and expensive.
The patterning steps (Fig.13(left)) involve (a) the appli-cation (and alignment, if needed) of the shadow mask on the membrane, (b) the deposition of the metal film and (c) the removal of the mask. The main advantage of this technique is its rapidity due to the possibility of re-using the masks. The disadvantages include the contact of the mask with the dielectric membrane, and the shape limitation (i.e. the dif-ficulty of patterning closed paths such as a ring). Lacour
Fig. 13 The three principal methods to pattern metallic thin films: Shadow masking (left), blanket deposition followed by a photolithography and a metal etch (center), and lift-off (right)
et al. have used shadow mask made in a polyimide foil to define conductive stripes on a silicone substrate [63], and the Danfoss Polypower metallization process can be used in conjunction with a roll-to-roll shadow mask [59].
Metal etching Photolithography combined with metal etching is another possibility (Fig. 13(center)). It consists of depositing a blanket metal electrode on the elastomer (a). A thin photoresist layer is then applied on the metal and patterned to selectively protect the metal layer (b). This step is followed by the etching of the exposed metal (c), and finally the photoresist is stripped from the metal (d). This process, which is the standard patterning method used in the integrated-circuit (IC) industry, has been success-fully demonstrated to pattern metallic electrodes on elas-tomeric membranes for DEA applications by several authors [9,27,50,53,64]. Compared to the shadow mask process, any shape can be realized with photolithography, and fea-tures with size down to 5 µm have been demonstrated on soft polymers [9]. However, the number of steps and the re-quired processing time are higher than for the shadow mask process.
Lift-off Another patterning possibility using photolithog-raphy and requiring a reduced amount of steps is the lift-off process (Fig.13(right)). A layer of photoresist is applied and patterned on the elastomer membrane (a). Unlike the metal-etch case, the exposed zones of the membrane after the development of the resist are the regions that need to be coated. The metal is then deposited over the patterned pho-toresist (b). Finally, the metal-covered resist is stripped in solvent, leaving the patterned metallic electrode on the elas-tomer (c). The lift-off process has been successfully used to pattern metallic layers on soft elastomers [55,65–67].
Photolithography on silicone membranes Conventional recipes for photolithography involve pre- and post-bakes of the photoresist at high temperature, which must be avoided with a silicone substrate because of the risk of generating cracks in the metallic and photoresist layers due to the differ-ence in the coefficient of thermal expansion (CTE) between the elastomeric membrane and the photoresist. A specific process for photolithography on soft elastomer must be de-veloped, as stated by Pelrine et al. without giving details regarding their method [9]. A few authors have published specific information on their photolithographic process on silicones, whose focus is mainly aimed towards avoiding the formation of cracks in the resist caused by the expansion of the silicone layer. For example, Pimpin et al. have placed the photoresist-coated substrate in a desiccator at room temper-ature for 12 hours in order to remove the solvent from the re-sist, instead of the traditional hot-plate or oven baking [50]. Some research groups have used the negative photoresist
SU-8 to pattern structures on a silicone layer, because it is stronger and has a much larger coefficient of thermal expan-sion compared to conventional resists, thus avoiding the for-mation of cracks during the process [65–67]. Because SU-8 is an epoxy, it can therefore not be stripped in solvents after cross-linking, and the SU-8 mask must therefore be mechan-ically peeled off from the substrate. This imposes limitation on the mask design, as the crosslinked SU-8 must form a continuous single part. Guo and DeWeerth have solved this limitation by using an intermediate sacrificial layer: a water-soluble poly(acrylic acid) (PAA) was coated directly on the PDMS, the photopatternable SU-8 being then spun on top of it. The PAA layer is dissolved in water as a last step, to allow the stripping of the SU-8 mask [65].
In addition to CTE mismatch-induced cracking of the photoresist layer, the proper wetting of the silicone surface by the photoresist and the adhesion thereof represent another challenge, due to the extremely low surface energy of sili-cone. Consequently, standard novolak-based resists have a tendency to dewet when directly applied on silicone, which is the case in the lift-off process. To address this issue, Maghribi et al. have used oxygen plasma activation (1 min at 100 watts with an oxygen flow of 300 sccm) just before spin-coating the resist, in order to increase the surface en-ergy of PDMS and ensure proper wetting by the photoresist [56]. Diebold and Clarke present another approach with the use of a mixture of Polydimethylglutarimide (PMGI) and standard Novolak resist, as PMGI was shown to present a good adhesion on silicone [68].
4.5 Interconnects
Depending on the targeted application and required output strain, metal thin-film electrodes may have too large a stiff-ening effect on the elastomer membrane, even when ap-plied with one of the methods described above. There are therefore situations for which carbon electrodes would be more suited. However, one of the main advantage of metal-lic electrodes are their high conductivity and ease of pattern-ing down to the micrometer scale. Even for actuators which would require a highly compliant active zone made with a carbon-based electrode, compliant metal thin films can be used as interconnect to bring the electrical signals from the actuator’s frame to the active zone. Long and narrow tracks with a low electric resistance can be patterned on the dielec-tric and used as soft interconnects, which is not feasible with carbon electrodes: in addition to the difficulty of patterning small structures, a narrow and long interconnect made with a material having a surface resistance in the tens of k/ range is not realistic.
The important parameters of interconnects are (a) a high conductivity and (b) the ability to be repeatably stretched without damage. The stiffening effect on the membrane is
of secondary importance. In addition to the three types of thin-film electrode presented above (patterned electrodes, buckled electrode and corrugated electrodes), there are a few other technologies that can be used for interconnects, while having a too large stiffening impact to be used as ac-tive electrode in a DEA. For example, several authors reports on large uniaxial strains (up to 30 %) obtained with metallic thin films deposited on a flat elastomeric substrate, greatly exceeding the usual 3–4 % yield strain of metals [64,69]. The authors have discovered that microcracks are created in the strained metallic tracks. As these cracks do not propa-gate completely across the track, a conduction path is con-served up to large strains. It is difficult to evaluate the impor-tance of the deposition parameters and of the stretching pro-file (very small increments with long pauses between each steps [64,69]) on the elongation at break for metallic thin films. As a comparison, we have previously strained sput-tered gold layers on PDMS at a constant speed and observed loss of conductivity for strains as low as 3 % [70].
5 Novel techniques for compliant electrodes
In addition to the two mainstream types of electrode (car-bon and metal thin films) described in the previous sections, some research groups have worked towards developing new types of compliant electrode for DEAs, in order to progress towards the perfect electrodes for DEAs combining compli-ance with ease of patterning, or to develop new useful fea-tures such as self-clearing in case of dielectric breakdown. This section presents some of the emerging and exotic meth-ods to make compliant electrodes for DEAs and pinpoints the advantages compared to the more conventional methods.
5.1 Implantation of metallic nanoclusters
Carbon powder electrodes are extremely compliant because they are formed by nano/micro particles that can slide rela-tive to each other when the elastomeric membrane on which they reside is stretched. However, their adhesion to the mem-brane is extremely weak, making them messy and unreli-able. On the other hand, metal thin films present a much better adhesion but lack compliance. Implantation of metal-lic nanoclusters at low energy takes the best of the two worlds by creating metallic nanoparticles in the first tens of nanometers below the elastomer surface (Fig.14(a)–(b)). These metallic clusters can move relative to each other, thus forming an electrode much more compliant than a plain thin film, and because they are located inside the elastomer ma-trix, the adhesion and resistance of these electrodes is excel-lent.
Two different implantation methods leading to the for-mation of metallic nanoclusters into PDMS have been pre-sented in the literature: Filtered Cathodic Vacuum Arc
Fig. 14 (a) TEM cross-section of Au nanoclusters implanted by FCVA into PDMS with an energy of 10 keV and a dose of 1.5× 1016at cm−2. From [74] ©2010 Elsevier ltd. Reproduced with permission. (b) TEM cross-section of a Au/PDMS nano-composite obtained by SCBI for an equivalent thickness (i.e. thickness obtained when depositing on a hard substrate) of 35 nm (exact dose unknown). From [75], ©2011 WILEY-VCH Verlag GmbH & Co. Reproduced with permission. (c) Schematic representation of the FCVA implan-tation process: the high density plasma is created by a high-voltage pulse that vaporizes the cathode material. A 90◦bent solenoid filters the plasma by trapping the large, heavy macroparticles. From [74], ©2010 Elesevier ltd. Reproduced with permission
(FCVA) implantation [70–74] and Supersonic Cluster Beam Implantation (SCBI) [75].
Fig. 15 Change of resistance versus strain for a gold-sputtered and a gold-implanted strip up to the damage threshold. The sputtered layer can only sustain a few percent strain whereas the implanted strip can be stretched up to 175 % before losing conductivity. From [70], ©2009 WILEY-VCH Verlag GmbH & Co. Reproduced with permission
Filtered cathodic vacuum arc In FCVA, a plasma is gener-ated between the cathode (the metal to be implanted) and a counter electrode (anode) by a high-voltage pulse that vapor-izes the cathode material (Fig.14(c)). The plasma, consist-ing of electrons, metal ions and large, undesirable particles is then filtered by a magnetic filter to remove the macropar-ticles [76]. The metal cations are finally accelerated towards the target by an electric field. Our group has worked exten-sively on low energy (2–10 keV) FCVA implantation of ti-tanium, palladium and gold ions into PDMS targeted at the fabrication of compliant electrodes for DEAs. Our investi-gation of gold-implanted PDMS have shown a percolation threshold of about 1.5× 1015at/cm2, leading to reasonably low surface resistance (∼1 k/), large stretching capa-bilities while remaining conductive and low impact on the Young modulus of the silicone membrane [70]. Conductive strips were implanted with gold on a silicone membrane and stretched. The resistance versus strain was measured with a four-points setup. Strains of up to 175 % were ob-served before the electrode was damaged, which is much larger than a gold-sputtered strip of identical size (Fig.15). The implanted layer extended from the surface to a depth of 20–30 nm depending on acceleration energy. The Young modulus of this ultra-thin implanted layer was measured to be 0.2–5 GPa depending on energy and dose, much lower than the Young modulus of bulk gold [74]. When used to make electrodes on a 25 µm PDMS membrane whose initial Young’s modulus is 1 MPa, ion implantation creates a 100 % relative increase of the stiffness, which is much lower than the impact of a plain evaporated thin film of comparable thickness. Gold-implanted tracks were cyclically stretched to 20 % without degradation [70].
Circular buckling DEAs with FCVA gold-implanted electrodes were fabricated and characterized [73,77]. They exhibited vertical displacement over diameter ratios up to
25 %, which is 4.5 times higher than a similar actuator made with patterned thin-film electrodes (cf. Sect.4). FCVA im-plantation is particularly interesting for the fabrication of compliant electrodes: First, the low energy limits the pen-etration depth of the ions into the elastomer, and allows percolation at low doses. The low energy also limits the collision-induced damage at the surface of the elastomer which contributes to stiffen the polymer (carbonization, bro-ken bonds, etc.). Secondly, the high ion flux allows obtain-ing a conductive layer in a short amount of implantation time (a few minutes).
Supersonic cluster beam implantation In plasma assisted SCBI, the metal is vaporized in the same manner as with FCVA, but the plasma is subsequently quenched by a pulse of inert gas and it condenses to form neutrally charged clus-ters. The clusters are then injected into the deposition cham-ber where they impact the target [78]. With this technique, neutral atoms are implanted and charging of the elastomer surface is therefore avoided. Corbelli et al. have used this method to create electrodes composed of conductive gold clusters into PDMS [75]. Even if the energy per atom is low (∼0.5 eV/at), whole clusters consisting of several thousand of atoms are reaching the target, and their kinetic energy is large enough to penetrate about 100 nm into the PDMS [75]. Using this method, Corbelli et al. obtained electrodes with a very low surface resistance (about 100 /) for an unspecified atom dose (180 minutes of implantation were necessary, which makes it a slow process). The conductive tracks were stretched up to 100 % while remaining conduc-tive. The upper limit was defined by the mechanical failure of the elastomer, and not by the electrode ceasing to con-duct. Cyclic stretching to 40 % for 50000 cycles showed a decreasing variation of the resistance change between the two extreme positions with the number of cycles. Because the authors of this study are not targeting DEA application specifically, they have not characterized the Young modu-lus increase caused by the implantation, but as the technique is very similar to FCVA implantation, a comparable impact should be expected. Indeed, TEM cross-sections of FCVA [74] and SCB [75] implanted silicone look similar, but with much larger clusters and penetration depth in the case of SCBI, probably due to the higher kinetic energy of the clus-ters for SCBI (Fig.14(a)–(b)).
Patterning methods for metallic nanoclusters electrodes Similar to metallic thin films, implanted metallic clusters can be patterned with the techniques presented in Sect.4.4: shadow masking and photolithography. In the latter case, only the lift-off process can be used. It is indeed not possi-ble to selectively remove the clusters with a metal etchant, as they are implanted and thus below the surface. Us-ing shadow maskUs-ing, our group has fabricated an array of
100×100 µm2actuators for single cell stretching that are ca-pable of deformation up to 80 % actuation strains [79], thus demonstrating that implanted electrodes can be patterned on the micron-scale while being much more compliant and presenting a better adhesion to the substrate compared to their thin-film cousins. The process of creating metal clus-ters electrodes is similar to that of depositing metal thin film by evaporation or sputtering in terms of involved equip-ment (vacuum chamber) and process flow. However, metal cluster electrodes have the advantage of being intrinsically compliant without the need of a special membrane micro-structure (cf. Sect.4.3), prestretching (cf. Sect.4.2) or pat-terning of the electrodes (cf. Sect. 4.1), as necessary for metal thin films. Furthermore, unlike corrugated and buck-led electrodes, metal cluster electrodes are compliant in the two in-plane directions, hence placing no restriction on the strain direction. Implanted metallic cluster electrodes are thus representing a progress towards metallic compliant and patternable electrodes for DEAs.
5.2 Photopatternable electrodes
Patterning of the compliant electrodes presented up to now was always made extrinsically, with the help of stencils or photoresists. However, intrinsically patternable compliant electrodes are very desirable, for they can greatly simplify the structuring process. Delille, Urdaneta et al. have reported on photosensitive electrodes based on platinum salts mixed in a photosensitive elastomer (Loctite 3108) [80–82]. The ingenuity of the process resides in the use of a filler (Pt salt) which is transparent to UV light. Indeed, previous attempts to produce a photopatternable conductive elastomer by mix-ing conductive fillers in a photosensitive elastomeric matrix were not conclusive because of the absorption and the scat-tering by the filler particles of the UV light needed for the cross-linking [81].
The elastomer-salt composite is applied on a substrate and exposed to UV light through a mask to locally induce cross-linking of the loaded elastomer, and the uncrosslinked elastomer is then subsequently flushed away in acetone. The composite is then immersed into an aqueous reducing solu-tion, causing the Pt salt to diffuse to the surface, reducing to the metal at the polymer/liquid interface (Fig.16). The per-colation threshold is observed for 6–7 % vol Pt salt at which point the surface resistance of the electrode is∼200 /, and it drops down to 2.2 / for 15 % vol [80]. The Young modulus of the composite is shown to be unaffected by the quantity of filler and remains close to that of pure Loctite 3108 (10 MPa) [82]. However, the stiffness of the photosen-sitive elastomer used in this study is one order of magnitude higher than elastomers commonly used for DEAs, and di-rect comparison with the stiffening effect of the other kind of compliant electrodes is not possible. For an application as
Fig. 16 Photopatternable and stretchable platium salt electrodes on elastomer [80] ©2007 WILEY-VCH Verlag GmbH & Co. Reproduced with permission
compliant electrodes for DEA, the Loctite 3108 elastomer should be replaced by an UV-sensitive low-durometer sili-cone, such as the SEMICOSIL 945 UV (Wacker, Germany). Patterned structures down to 350 µm were obtained with the Loctite 3108/Pt salt compound, whereas a resolution of 310 µm was observed for pure Loctite 3108. The slight de-crease in resolution in presence of the salt is attributed to light scattering [80]. These electrodes can be stretched up to 150 % before losing conductivity and have been cyclically stretched up to 30 %, at which point mechanical failure oc-curs after about 1000 stretching cycles. For smaller strains, a degradation is observed during the first hundreds of cycles, but it then stabilizes. The mechanical and electrical prop-erties of these electrodes depend on the content of Pt salt, which was varied between 9 and 14 %.
The downside of this pioneering method, although intrin-sically patternable, is its relatively poor resolution compared to the previously mentioned techniques, such as stencil, or photolithography.
5.3 Self-clearing electrodes
One of the principal failure mode of DEAs is electrical breakdown of the dielectric membrane due to an exces-sive electric field applied between the electrodes. The global electric field is equal to the applied voltage divided by the average thickness of the membrane. However, the electric field can be locally higher, for example because of defects in the membrane. When breakdown occurs, a current flows through the membrane at the location of the defect. The consequence of breakdown is usually a permanent conduc-tive carbon track being created through the membrane, ef-fectively short-circuiting the electrodes, and thus preventing the actuator to charge and move. Depending on the current output capability of the source used to drive the actuator, breakdown can also cause puncture of the membrane, which
